Latham & Watkins buys Nvidia servers to run AI systems in house

In short: Latham & Watkins has bought Nvidia GPU servers and is tailoring AI models to run inside the firm, rather than relying only on outside AI providers.

What happened

Latham & Watkins, one of the largest US law firms, has purchased several servers that contain Nvidia GPUs, which are powerful computer chips often used to run AI systems. The firm is using this equipment to build and operate its own AI setup.

Instead of using only services from companies like OpenAI or Anthropic, Latham is fine-tuning “open-weight” AI models, meaning models that can be downloaded and adjusted (like buying a recipe you can change, instead of ordering a finished meal). The firm said this lets it keep very sensitive client information within its own systems, rather than sending it to a cloud provider, which is an outside company that runs software on its own computers.

The firm leases space in a data center that is locked and only accessible by its staff. It has a large internal tech team, and said it employs more than 900 technology specialists. Latham declined to say how much it is spending, but the Financial Times reported that buying, running, and securing this kind of system can cost tens of millions of dollars a year.

Why it matters

Legal work often involves reviewing huge numbers of documents and repeating similar steps, so AI can speed up many tasks. At the same time, clients like banks are pushing law firms to lower fees when work takes less time. Latham’s move shows a growing tradeoff: running AI in house can offer more privacy and more control over future pricing, but it also means the firm must handle security and the high ongoing costs itself.

Source: Financial Times